mjhjmtu

For example:

22111155
44788
56667775
aannmmmm88
66h1122
PPDd88

The output should be:

44788 # 7 is not repeated
66h1122 # letter h is not repeated 
PPDd88 # letters D or d are not repeated

Delete the line that all its characters are repeated at least two times in the same line.

If there was any character not repeated in the same line so print that line only.

With perl:

perl -ne 'my %count;
 $count$_++ for /./g;
 print if grep $_ == 1 values %count'

With sed:

sed '
/./!d;h;s/$/
/
:1
 s/(.)(.*)1(.*n)/231/
 s/(.)1*(.*n.*1)/2/
t1
/^n/d;g'

We split the pattern space into two lines. Characters that are duplicated are moved to the second line in a loop. We print the record if at the end, the first line is not empty.

perl one-liner: removes all pairs of characters, print the line if there are characters left over.

perl -lne '($copy = $_) =~ s/(.)1//g; print if $copy' file

As you say, the above is wrong: it will incorrectly print "56667775" because that answer only looks at pairs of characters. Look to Stéphane's answer for correctness.

There is a command that extract only non-repeated words in a list:

$ printf '%sn' one one two | uniq -u
two

You could divide every character in a word on each line and use uniq:

$ echo "1122e4455" | grep -o . | sort | uniq -u
e

All you have to do is loop for all the words to test and if the command above doesn't have any output, print the tested line.

Adaption of the solution to your recent problem:

awk '
 split ("", N) # delete N array
 L = 0 # reset boolean L used for print decision
 for (i=1; i<=length; i++) N[substr($0, i, 1)]++ # calculate count of characters
 for (n in N) if (N[n] < 2) L = 1 # for non-duplicate chars: set print decision
 break # and quit the for loop
 
 
L # print if non-duplicate chars exist
' file
44788
66h1122
PPDd88

sed -e '
 /n/!h;s/^/n/;
 /^n$/d
 /^n(.).*1/!g;b;
 :b;s/^(n(.).*)2/1/;tb
 s/n./n/;s/^/n/;D
'

Explanation:

• Place a marker n at the beginning of the pattern ,which travels to the right during the process.


• We setup an infinite loop and provide for two exits inside the loop.


• One, if during the process the whole string ins emptied leaving only the marker, we exit knowing that this string comprised all duplicate stuff.


• Two, if during the process we find that the first element in the string is non-repeated. Implies, at least one non-duplicate located. Hence this line needs printing. We pull out the original which we had stored away in the hold area prior to beginning the process.


• The alternative is that the first element has at least a duplicate somewhere in the string which we go about in t-loop totally purging that element's presence from the string.

Another way is usingPerl as shown:

perl -lne '
 my $s = $_;
 s/Q$1E//g while $s =~ /(.)(?=.*?1)/g;
 print $s if /./;
'

Here we make a copy of the line, and continuously look for duplicated elements and keep stripping away these from the original line. After the while loop ends, we print the copy provided something remains in the line (meaning, the non-duplicated stuff was left behind).